[go: up one dir, main page]
More Web Proxy on the site http://driver.im/

EP1022632A1 - Überwachungs- und Diagnoseeinrichtung mit neuronalem Netzwerk zur Modellierung der normalen oder anormalen Funktionsfähigkeit eines elektrischen Gerätes - Google Patents

Überwachungs- und Diagnoseeinrichtung mit neuronalem Netzwerk zur Modellierung der normalen oder anormalen Funktionsfähigkeit eines elektrischen Gerätes Download PDF

Info

Publication number
EP1022632A1
EP1022632A1 EP99200157A EP99200157A EP1022632A1 EP 1022632 A1 EP1022632 A1 EP 1022632A1 EP 99200157 A EP99200157 A EP 99200157A EP 99200157 A EP99200157 A EP 99200157A EP 1022632 A1 EP1022632 A1 EP 1022632A1
Authority
EP
European Patent Office
Prior art keywords
diagnostic apparatus
values
monitoring diagnostic
fact
neural network
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Withdrawn
Application number
EP99200157A
Other languages
English (en)
French (fr)
Inventor
Claudio Franchi
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
ABB Research Ltd Switzerland
ABB Research Ltd Sweden
Original Assignee
ABB Research Ltd Switzerland
ABB Research Ltd Sweden
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by ABB Research Ltd Switzerland, ABB Research Ltd Sweden filed Critical ABB Research Ltd Switzerland
Priority to EP99200157A priority Critical patent/EP1022632A1/de
Priority to JP10865A priority patent/JP2000215766A/ja
Publication of EP1022632A1 publication Critical patent/EP1022632A1/de
Withdrawn legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B23/00—Testing or monitoring of control systems or parts thereof
    • G05B23/02—Electric testing or monitoring
    • G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
    • G05B23/0218—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults
    • G05B23/0224—Process history based detection method, e.g. whereby history implies the availability of large amounts of data
    • G05B23/024—Quantitative history assessment, e.g. mathematical relationships between available data; Functions therefor; Principal component analysis [PCA]; Partial least square [PLS]; Statistical classifiers, e.g. Bayesian networks, linear regression or correlation analysis; Neural networks
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B23/00—Testing or monitoring of control systems or parts thereof
    • G05B23/02—Electric testing or monitoring
    • G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
    • G05B23/0259—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterized by the response to fault detection
    • G05B23/0275—Fault isolation and identification, e.g. classify fault; estimate cause or root of failure
    • G05B23/0281—Quantitative, e.g. mathematical distance; Clustering; Neural networks; Statistical analysis

Definitions

  • the present invention relates to a monitoring diagnostic apparatus for detecting the normal or abnormal functionality of electrical devices such as circuit breakers, switches, disconnecting switches, oil-insulated electrical equipment, vacuum-insulated electrical equipment, air-insulated electrical equipment, solid insulated electrical equipment.
  • the present invention relates to a monitoring diagnostic apparatus which allows to detect the main causes of failures of an electrical device through the use of a neural network to model the normal or abnormal functionality of the electrical device and to identify the cause of said abnormal functionality, if present.
  • monitoring diagnostic apparatuses for detecting faults in electrical devices such as circuit breakers, switches, etc.
  • the conventional monitoring apparatuses use a large amount of different sensing devices for fault monitoring. Such sensing devices are connected to means for making a comparison between the signals output by said sensors and predefined threshold values. Due to their intrinsic structure, the conventional monitoring diagnostic apparatuses allow to detect the presence of an abnormal functionality of the device (for example when the intensity of the output signal is greater or smaller than the threshold value) but do not provide any aid to identify the cause of said abnormal functionality. In addition, the conventional monitoring diagnostic apparatuses cannot adapt themselves to changes of the monitored device conditions depending, for example, on changes of the type of istallation.
  • FIG 1 it is shown a schematic representation of a neural network based monitoring diagnostic apparatus of the state of the art, as proposed in the US patent N" 5,419,197.
  • a vibration sensor 106 is positioned outside the tank 105 of a circuit breaker for measuring an acceleration due to vibrations transmitted to the tank 105 by pressure waves originated by partial discharges inside the tank.
  • the sensor 106 provides a signal 107, proportional to accelleration, to a signal processing means 108 which processes said signal 107 and provides an input to a neural network 114.
  • Signal processing means 108 performs some processing steps which consist in a Fast Fourier Transform 109 of the signal, an operation of averaging 110 in the domain of frequency and an operation of normalization 113 in order to obtain a collection of values to put into the neural network.
  • the neural network provides output data to a diagnostic display 118.
  • the apparatus illustrated in figure 1 is provided for the detection of the presence of partial discharges inside the tank 105.
  • the presence of partial discharges generates pressure waves which induce the tank 105 to vibrate.
  • partial discharges represent a cause of abnormal functionality of the electrical device in a little number of cases. More often, abnormal functionality is due to failures related to the driving mechanism of the electrical device. Such failures cannot be detected by the apparatuses illustrated in figure 1 due to the location of the accelleration sensor that, being placed outside the tank, cannot detect the vibrations generated by failing parts of the driving mechanism.
  • the signal processing means considered by the monitoring diagnostic apparatuses of the state of the art, utilize very complicated alghoritms that require a large amount of computations and therefore expensive processing modules are needed.
  • performing a Fast Fourier Transform implies the use of algorithms whose complexity is proportional to the square of the amount of input data.
  • Main aim of the present invention is to provide a monitoring diagnostic apparatus which allows to detect and identify the main causes of abnormal functionality of an electrical devices such a circuit breakers, switches, disconnecting switches, oil-insulated electrical equipment, vacuum-isulated electrical equipment, air-insulated electrical equipment, solid insulated electrical equipment.
  • an other object of the present invention is to provide a monitoring diagnostic apparatus which allows to characterize the operation of the driving mechanism of electrical devices.
  • Further object of the present invention is to provide a monitoring diagnostic apparatus which uses a neural network to identify causes of abnormal functionality.
  • An other object of the present invention is to provide a monitoring diagnostic apparatus which uses a relatively low number of sensing devices.
  • An other objectct of the present invention is to provide a monitoring diagnostic apparatus which uses simple alghoritms for processing the signals output by such sensors.
  • a monitoring diagnostic apparatus for detecting the normal or abnormal functionality of a driving mechanism of an electrical device comprising:
  • the apparatus as in the present invention, is characterized by the fact that:
  • a sensor 6 is placed inside the tank 5 of the electrical device 3, for example a high or medium voltage circuit breaker, for detecting a quantity characterizing the behaviour of the driving mechanism 1 of the electrical device 3.
  • the vibrations of the movable parts of the driving mechanism can be monitored, if the sensor 6 is placed aboard the driving mechanism 1.
  • the sensor 6 can be an acceleration sensor. The placement of the sensor 6 is particularly advantageous because it allows to identify the main causes of failure of the electrical device 3.
  • the sensor 6 provides a signal 7 representing the detected quantity.
  • the signal 7 is put into signal processing means 12 which elaborates the signal 7 and outputs a collection of values 13 to put into a neural network 14.
  • the neural network 14 outputs signals of normal or abnormal functionality of the driving mechanism 1 to a diagnostic display 18, indicating the type of fault, if present.
  • a sampling of the signal 7 is performed and a time window, comprising a predefined number of values sampled from the signal 7, is calculated.
  • a shift value is calculated. This shift value is necessary for obtaining, for example through a simple product, a sequence of time windows covering the time domain of the signal 7.
  • a windowed average is calculated. This implies that the average of the values comprised in each window is calculated obtaining a collection of values 13 to put into the neural network 14. As it will be illustrated hereinafter, this calculation can be performed in different ways.
  • the elaborations performed in step 22 are particularly advantageous because they allow to reduce hugely the magnitude of the amount of data to put into the neural network for the classification of the faults detected.
  • the neural network 14 is divided in an input layer 15, an hidden layer 16 and an output layer 17.
  • Each layer contains a plurality of neural elements.
  • the input layer contains a neural element for every value of the collection 13.
  • Each element of the input layer 15 is connected to each element of the hidden layer by a first type of connectivity weights 150, while each element of the hidden layer 16 is connected to each element of the output layer 17 through a second type of connectivity weights 160.
  • the output level 17 of the neural network provides logical values able to activate the proper messages for the diagnostic display 18. The calculation of such connectivity weights 150 and 160 is illustrated herebelow.
  • FIG 4 a schematic representation of the calculations performed by processing steps of figure 3 is presented.
  • the signal 7 is then sampled at step 20 in figure 3 and the sampling period is indicated in figure 4 as Ts.
  • the sampling period is predefined since the time domain of the signal 7 can be expressed as (M xTs), M being a predefined number.
  • the extension of a time window is thus given by (W xTs), where W is a predefined number of sampled values to be contained.
  • W xTs The shift value necessary to create a sequence of time windows along the time domain of signal 7 is given by (S xTs), where S is a predefined number.
  • the windowed average for each window can be performed in different ways.
  • the collection of values p k is the collection 13 described in figure 3.
  • two collections of values for inputting the neural network can be provided. One collection is referred to positive sampled values while the other one is referred to negative sampled values.
  • the two collections can be given by the following equations ( 2 ) and ( 3 ):
  • the calculation of the collection of the input values 13 can in practise be perfomed by a simple iterative software loop. This is particularly advantageous because it allows to mantain linear the cardinality order of the alghoritm. This fact means that very simple processing resources are required.
  • FIG 5 a schematic view of the neural network is presented. In a preferred embodiment all the elaborations are supposed to happen at the hidden level 16.
  • the input layer 15 has many neural elements as the number of values provided by the collection 13 of figure 2. All the neural elements of the input layer 15 are linked to the neural elements of the hidden layer 16 through the connectivity weights 150 while all the elements of the hidden layer 16 are connected to the neural elements of the output layer 17 through the connectivity weights 160.
  • Each neural element of the output layer is associated to a type of fault or to the normal functioning condition. The message relative to a particular type of fault is actived when the associated real value assumes a suitable level.
  • the neural network is submitted to a learning process which mainly consists in the evaluation of the connecitvity weights 150 and 160 of figure 2.
  • the h th signal outputted by an element of the hidden layer is given by the following equation ( 4 ): where N is the number of neural elements of the input layer, H is the number of neural elements of the hidden layer.
  • the signal put into the o th neural element of the output layer is given by the following equation ( 5 ): where O is the number of neural elements of the output layer, H is the number of neural elements of the hidden layer.
  • the signal Vo is a digital signal.
  • the connectivity weights 150 are a sigmoid function depending on the input values from the collection 13, while the connectivity weights 160 are costant values.
  • the learning process comprises the following steps:
  • the calculation can be realized by known procedures: for example a Maximum Gradient Descent Procedure (MGDP) can be easily used.
  • MGDP Maximum Gradient Descent Procedure
  • the monitoring diagnostic apparatus as in the present invention has shown to fulfill all the above mentioned aim. Infact, it has been demonstred in practive that the placement of the sensor inside the tank of the electrical device allows advantageosly to identify the most important fault conditions. This gives huge advantages when service operations has to be performed. Moreover the simple structure of the signal processing means allows to obtain relatively low cost for the installation and the operation of the monitoring diagnostic apparatus.
  • a neural network can easily allow to identify the abnormal or normal functionality of the electrical device.
  • the apparatus as in the present invention can be easily adapted to different installation types thanks to simply training the network using different "a priori" values.
  • circuit breaker has been taken as example of electrical device.
  • objects of the embodiments are not limited to a circuit breaker but can also be applied to other electrical equipment such as switches, disconnecting switches, oil-insulated electrical equipment, vacuum-isulated electrical equipment, air-insulated electrical equipment, solid insulated electrical equipment.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Automation & Control Theory (AREA)
  • Mathematical Physics (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Algebra (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Probability & Statistics with Applications (AREA)
  • Pure & Applied Mathematics (AREA)
  • Testing Electric Properties And Detecting Electric Faults (AREA)
  • Testing And Monitoring For Control Systems (AREA)
  • Driving Mechanisms And Operating Circuits Of Arc-Extinguishing High-Tension Switches (AREA)
EP99200157A 1999-01-21 1999-01-21 Überwachungs- und Diagnoseeinrichtung mit neuronalem Netzwerk zur Modellierung der normalen oder anormalen Funktionsfähigkeit eines elektrischen Gerätes Withdrawn EP1022632A1 (de)

Priority Applications (2)

Application Number Priority Date Filing Date Title
EP99200157A EP1022632A1 (de) 1999-01-21 1999-01-21 Überwachungs- und Diagnoseeinrichtung mit neuronalem Netzwerk zur Modellierung der normalen oder anormalen Funktionsfähigkeit eines elektrischen Gerätes
JP10865A JP2000215766A (ja) 1999-01-21 2000-01-19 モニタリング診断装置

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
EP99200157A EP1022632A1 (de) 1999-01-21 1999-01-21 Überwachungs- und Diagnoseeinrichtung mit neuronalem Netzwerk zur Modellierung der normalen oder anormalen Funktionsfähigkeit eines elektrischen Gerätes

Publications (1)

Publication Number Publication Date
EP1022632A1 true EP1022632A1 (de) 2000-07-26

Family

ID=8239814

Family Applications (1)

Application Number Title Priority Date Filing Date
EP99200157A Withdrawn EP1022632A1 (de) 1999-01-21 1999-01-21 Überwachungs- und Diagnoseeinrichtung mit neuronalem Netzwerk zur Modellierung der normalen oder anormalen Funktionsfähigkeit eines elektrischen Gerätes

Country Status (2)

Country Link
EP (1) EP1022632A1 (de)
JP (1) JP2000215766A (de)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8836536B2 (en) 2011-07-29 2014-09-16 Hewlett-Packard Development Company, L. P. Device characterization system and methods
WO2016033247A2 (en) 2014-08-26 2016-03-03 Mtelligence Corporation Population-based learning with deep belief networks
WO2019075612A1 (en) * 2017-10-16 2019-04-25 Abb Schweiz Ag CIRCUIT BREAKER MONITORING METHOD AND APPARATUS AND INTERNET OF OBJECTS USING THE SAME
CN109784478A (zh) * 2019-01-16 2019-05-21 江苏圣通电力新能源科技有限公司 一种基于bp神经网络的配电系统传感器故障诊断方法
WO2019187297A1 (en) * 2018-03-28 2019-10-03 Mitsubishi Electric Corporation Apparatus and method for controlling system
CN113537288A (zh) * 2021-06-16 2021-10-22 华北电力大学 基于样本信号的修改对目标模型进行训练的方法及系统

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5419197A (en) * 1992-06-02 1995-05-30 Mitsubishi Denki Kabushiki Kaisha Monitoring diagnostic apparatus using neural network
US5566092A (en) * 1993-12-30 1996-10-15 Caterpillar Inc. Machine fault diagnostics system and method

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5419197A (en) * 1992-06-02 1995-05-30 Mitsubishi Denki Kabushiki Kaisha Monitoring diagnostic apparatus using neural network
US5566092A (en) * 1993-12-30 1996-10-15 Caterpillar Inc. Machine fault diagnostics system and method

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8836536B2 (en) 2011-07-29 2014-09-16 Hewlett-Packard Development Company, L. P. Device characterization system and methods
WO2016033247A2 (en) 2014-08-26 2016-03-03 Mtelligence Corporation Population-based learning with deep belief networks
EP3183622A4 (de) * 2014-08-26 2017-11-01 Mtelligence Corporation Populationsbasiertes lernen mit deep-belief-netzwerk
WO2019075612A1 (en) * 2017-10-16 2019-04-25 Abb Schweiz Ag CIRCUIT BREAKER MONITORING METHOD AND APPARATUS AND INTERNET OF OBJECTS USING THE SAME
US11656279B2 (en) 2017-10-16 2023-05-23 Hitachi Energy Switzerland Ag Method for monitoring circuit breaker and apparatus and internet of things using the same
WO2019187297A1 (en) * 2018-03-28 2019-10-03 Mitsubishi Electric Corporation Apparatus and method for controlling system
JP2021509995A (ja) * 2018-03-28 2021-04-08 三菱電機株式会社 システムを制御する装置及び方法
CN109784478A (zh) * 2019-01-16 2019-05-21 江苏圣通电力新能源科技有限公司 一种基于bp神经网络的配电系统传感器故障诊断方法
CN113537288A (zh) * 2021-06-16 2021-10-22 华北电力大学 基于样本信号的修改对目标模型进行训练的方法及系统
CN113537288B (zh) * 2021-06-16 2024-04-05 华北电力大学 基于样本信号的修改对目标模型进行训练的方法及系统

Also Published As

Publication number Publication date
JP2000215766A (ja) 2000-08-04

Similar Documents

Publication Publication Date Title
KR101874286B1 (ko) 전력설비 모니터링 및 진단 시스템
KR100755955B1 (ko) 수배전 설비 고장 진단 시스템
US9910093B2 (en) Generator neutral ground monitoring device utilizing direct current component measurement and analysis
US5903158A (en) Monitoring of internal partial discharges in a power transformer
CN101688820B (zh) 用于使用定子电流噪音消除的轴承故障检测的系统和方法
US5305235A (en) Monitoring diagnosis device for electrical appliance
KR100719138B1 (ko) 수배전 설비 고장 진단 방법
JPH11118592A (ja) 機器異常診断装置およびそれを搭載したプラント装置
CA2346759A1 (en) Ingress monitoring system and method
CN111551352B (zh) 一种gis设备的断路器的状态的检测方法及系统
EP1022632A1 (de) Überwachungs- und Diagnoseeinrichtung mit neuronalem Netzwerk zur Modellierung der normalen oder anormalen Funktionsfähigkeit eines elektrischen Gerätes
KR20150118456A (ko) 감시 장치의 부분방전 진단 방법
US11249137B2 (en) Diagnostic method of switchgear and device thereof
EP2418502A2 (de) Systeme, Verfahren und Vorrichtungen Anschlussfehlerselbstüberwachung mit Gleichstrom-Vorstrom
KR100954665B1 (ko) 전력기기의 위험도 분석 시스템 및 그 방법
KR102231150B1 (ko) 부싱 진단 시스템
CN109375037A (zh) 一种中压船舶电力系统单相接地故障报警器设计方法
WO2009028875A2 (en) Method and apparatus for supervising integrated trip margin of facilities in industrial plant
KR20220166319A (ko) 전동기의 진단 장치
KR100543585B1 (ko) 부분방전 진단 장치 및 이를 이용한 데이터 전송방법
CN109427110A (zh) 一种诊断机械设备的方法
JP3172626B2 (ja) 高圧機器の部分放電検出方法
Takagi et al. Application of gas pressure sensor for fault location system in gas insulated substation
CN110553830A (zh) 一种金属封闭开关设备机械特性在线检测方法
JPH05176371A (ja) ガス圧力監視装置

Legal Events

Date Code Title Description
PUAI Public reference made under article 153(3) epc to a published international application that has entered the european phase

Free format text: ORIGINAL CODE: 0009012

AK Designated contracting states

Kind code of ref document: A1

Designated state(s): AT BE CH CY DE DK ES FI FR GB GR IE IT LI LU MC NL PT SE

AX Request for extension of the european patent

Free format text: AL;LT;LV;MK;RO;SI

AKX Designation fees paid
STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: THE APPLICATION IS DEEMED TO BE WITHDRAWN

18D Application deemed to be withdrawn

Effective date: 20010130

REG Reference to a national code

Ref country code: DE

Ref legal event code: 8566